agentleFS
vasceannievasceannie/slopgate/.agents/skills/oma-mobile/SKILL.md

oma-mobile

Skill1 starsChanged 51 days agoMITView on GitHub

Mobile specialist for Flutter, React Native, and Swift native iOS

We split this skill into the rules it gives an agent

13 rules. 10 are common, 0 are distinctive, and 3 are very specific to this project.

  • 10

    Common

    Most of the 4 edited versions we read keep these.

  • 0

    Distinctive

    This author's own choices. Worth a look.

  • 3

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Common:

    Test expectations for unit, widget, integration, or Maestro E2E coverage

    line 32
  2. Common:

    ACQUIRE: Read existing widgets/screens, providers/blocs, API clients, and tests.

    line 58
  3. Very specific:

    VERIFY: Run relevant unit/widget/integration/E2E checks.

    line 60
  4. Common:

    If business logic is complex, keep it in domain/data layers before presentation.

    line 64
  5. Common:

    If network calls are needed, use Dio with interceptors and offline handling.

    line 65
  6. Common:

    If a critical user flow changes, add or update Maestro E2E coverage.

    line 66
  7. Common:

    If backend contracts are missing, coordinate with backend/API work.

    line 67
  8. Common:

    If platform SDK or emulator is unavailable, report verification limits.

    line 70
  9. Common:

    If a permission or native capability is missing, add explicit platform configuration or document blocker.

    line 71
  10. Common:

    If tests fail, fix before handoff or report the failing check.

    line 72
  11. Very specific:

    All controllers disposed in dispose() method

    line 123
  12. Very specific:

    Use the platform transport with auth/retry/logging interception and offline handling: Flutter uses Dio, React Native uses axios behind …

    line 124
  13. Common:

    Use Maestro for E2E testing of critical user flows

    line 126

“Common” is measured: kept by at least 6 in 10 of the 4 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Mobile Agent - Cross-Platform Mobile Specialist

Scheduling

Goal

Build, modify, and verify cross-platform mobile application features with clean architecture, platform-appropriate UI, state management, performance, and E2E coverage.

Intent signature

  • User asks for mobile app, Flutter, Dart, React Native, iOS, Android, Riverpod, widgets, camera, GPS, push notifications, or offline-first work.
  • User needs native or cross-platform mobile behavior rather than web frontend work.

When to use

  • Building native mobile applications (iOS + Android)
  • Mobile-specific UI patterns
  • Platform features (camera, GPS, push notifications)
  • Offline-first architecture

When NOT to use

  • Web frontend -> use Frontend Agent
  • Backend APIs -> use Backend Agent

Expected inputs

  • Target screen, widget, feature, platform capability, or mobile flow
  • Existing app architecture, state management pattern, API contract, and platform constraints
  • Test expectations for unit, widget, integration, or Maestro E2E coverage

Expected outputs

  • Mobile code changes in domain, data, presentation, platform, or test files
  • UI aligned with Material Design 3 and iOS HIG as applicable
  • Verification results from mobile checks and critical-flow tests

Dependencies

  • Flutter/Dart or React Native toolchain as detected from the project
  • Riverpod/Bloc, Dio, platform SDKs, and Maestro where applicable
  • resources/execution-protocol.md, examples, snippets, checklist, and screen template

Control-flow features

  • Branches by platform, state management pattern, offline requirement, native permission, and test level
  • Reads and writes mobile codebase files
  • May call build, test, simulator, emulator, or E2E commands

Structural Flow

Entry

  1. Identify target platform(s), screen/feature, architecture layer, and state boundary.
  2. Inspect existing mobile patterns and dependencies.
  3. Determine test level and verification environment.

Scenes

  1. PREPARE: Load app architecture, platform constraints, and acceptance criteria.
  2. ACQUIRE: Read existing widgets/screens, providers/blocs, API clients, and tests.
  3. ACT: Implement mobile UI, state, platform integration, offline handling, and tests.
  4. VERIFY: Run relevant unit/widget/integration/E2E checks.
  5. FINALIZE: Report behavior, platforms covered, and verification results.

Transitions

  • If business logic is complex, keep it in domain/data layers before presentation.
  • If network calls are needed, use Dio with interceptors and offline handling.
  • If a critical user flow changes, add or update Maestro E2E coverage.
  • If backend contracts are missing, coordinate with backend/API work.

Failure and recovery

  • If platform SDK or emulator is unavailable, report verification limits.
  • If a permission or native capability is missing, add explicit platform configuration or document blocker.
  • If tests fail, fix before handoff or report the failing check.

Exit

  • Success: mobile feature works for target platforms and passes relevant checks.
  • Partial success: platform, simulator, dependency, or verification gaps are explicit.

Logical Operations

Actions

ActionSSL primitiveEvidence
Inspect mobile architectureREADDomain/data/presentation files
Select state and platform strategySELECTRiverpod/Bloc and platform constraints
Implement mobile codeWRITEWidgets, screens, providers, clients
Validate lifecycle and permissionsVALIDATEDispose, permissions, offline behavior
Call verification toolsCALL_TOOLTests, builds, Maestro
Report resultNOTIFYFinal summary

Tools and instruments

  • Flutter/Dart or React Native stack
  • Riverpod/Bloc, Dio, platform SDKs, Maestro
  • Unit, widget, integration, and E2E test commands

Canonical workflow path

rg --files
rg "Riverpod|Bloc|Dio|Widget|Maestro|dispose\\(|permission" .

Then run the project's mobile verification commands, typically unit/widget tests and Maestro E2E for critical flows.

Resource scope

ScopeResource target
CODEBASEMobile source, tests, platform config
LOCAL_FSTemplates, snippets, resources
PROCESSBuild, test, emulator, simulator, E2E commands
NETWORKBackend APIs when the feature integrates remotely

Preconditions

  • Target mobile feature and platform scope are identifiable.
  • Required SDKs, permissions, and API contracts are available or assumptions are stated.

Effects and side effects

  • Mutates mobile source, tests, and platform configuration.
  • May affect permissions, app lifecycle, offline data, or performance.

Guardrails

  1. Clean Architecture: domain -> data -> presentation
  2. Riverpod/Bloc for state management (no raw setState for complex logic)
  3. Material Design 3 (Android) + iOS HIG (iOS)
  4. All controllers disposed in dispose() method
  5. Use the platform transport with auth/retry/logging interception and offline handling: Flutter uses Dio, React Native uses axios behind TanStack Query, and Swift uses generated Client middleware.
  6. 60fps target; test on both platforms
  7. Use Maestro for E2E testing of critical user flows
  8. Swift native: SwiftUI + @MainActor @Observable view models (Observation framework, iOS 17+) — non-isolated VMs mutating observed state from a Task are a Swift 6 strict-concurrency error
  9. Swift native: use the generated Client from swift-openapi-generator — never hand-roll URLRequest/JSONDecoder for API calls
  10. Swift native: cache API responses at the Repository layer via a ResponseCache actor over hyperoslo/Cache — cache DECODED models (never HTTPBody), serve stale-while-revalidate on reads, invalidate keys on writes; view models depend on a protocol seam, not the concrete service (see variants/swift-ios/snippets.md §10)
  11. Swift native: follow App/Core/Features/Shared project layout
  12. Swift native: iOS Human Interface Guidelines for all UI decisions
  13. Swift native: XCTest or Swift Testing for units, XCUITest for critical flows; cancel work via structured .task {} (auto-cancels on disappear) — never in deinit, which is nonisolated and cannot touch @MainActor state under Swift 6 (isolated deinit requires Swift 6.2+)
  14. Swift native: restore edge swipe-back at the route layer — nav-bar-hidden screens (.toolbar(.hidden, for: .navigationBar)) lose it, so register push routes via a swipeBackDestination wrapper, not per-screen (see variants/swift-ios/snippets.md §9)
  15. Flutter: mandate a repository-layer offline-first cache (Drift) — read cached entities then revalidate (stale-while-revalidate), invalidate/refresh affected rows on every write; cache decoded entities at the data layer, never at the Dio transport (see variants/flutter/snippets.md §3, §10)
  16. React Native: server state goes through TanStack Query (the repository-layer cache) with explicit staleTime/gcTime — invalidate affected query keys on every mutation, persist the cache to MMKV for offline; screens consume query/mutation hooks, never call axios directly (see variants/react-native/snippets.md)

References

  • Execution steps (follow for the selected task): resources/execution-protocol.md
  • Code snippets (Swift): variants/swift-ios/snippets.md
  • Code snippets (Flutter): variants/flutter/snippets.md
  • Code snippets (React Native): variants/react-native/snippets.md
  • Checklist (run before handoff): resources/checklist.md
  • Error recovery: resources/error-playbook.md
  • Tech stack index (all platforms): resources/tech-stack.md
  • Tech stack (Swift): variants/swift-ios/tech-stack.md
  • Tech stack (Flutter): variants/flutter/tech-stack.md
  • Tech stack (React Native): variants/react-native/tech-stack.md
  • Screen template (Flutter): resources/screen-template.dart
  • Screen template (Swift): resources/screen-template.swift
  • Screen template (React Native): resources/screen-template.tsx
  • API service template (Swift): variants/swift-ios/api-template.swift
  • API service template (Flutter): variants/flutter/api-template.dart
  • API service template (React Native): variants/react-native/api-template.ts
  • Variant registry: variants/README.md
  • Context loading: ../_shared/core/context-loading.md
  • Clarification: ../_shared/core/clarification-protocol.md
  • Context budget: ../_shared/core/context-budget.md
  • Lessons learned: ../_shared/core/lessons-learned.md
  • Observability handoff: ../oma-observability/SKILL.md §Integrations — offline queuing, crash analytics, battery-aware sampling

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.